Keeping Up With the Coronavirus Variant Landscape

 BY DEREK LOWE

I'd like to recommend this excellent article by Olivia Goodhill at Stat for people wondering how you keep a vaccine up-to-date given the constant emergence of viral variants. It's an exclusive look at Pfizer's Pearl River site, which has that important but extremely demanding assignment, and it's a look at what really is the state of the art in commercial vaccine development. The usual warnings abou the complexity of human immunology apply, and how:

Studying Covid-19 has only emphasized how little we know. “A lot of times, when you’re working in this field, you’d look at the animal data and say this thing has a wimpy antibody response, let’s not go ahead with it,” said Roopchand. Data from the Phase 3 efficacy study of the Covid vaccine undermines that approach: Vaccinated participants have protections against the virus by day 12, at a time when there’s barely any antibody response. “That was the biggest surprise,” said Roopchand.

The good news is that these people (and many others around the world) are working frantically to characterize the viral mutations and the effectiveness of the current vaccines against them. That's basically an endless job right now. Millions upon millions of new coronavirus infections are happening, and each patient generates untold numbers of new viral particles, and each of the replications that generate each one of those is capable of throwing some new RNA variant into the mix. For any organism, fidelity in copying the genetic material is not so much an on-off switch as it is a dial marked "looser" and "tighter".

The amount that dial can turn and the levels it selects from are determined in a broad sense by evolutionary fitness. Many creatures (such as us!) have pretty extensive error-correcting machinery and are definitely on the "tighter" part of that rheostat dial. Viruses, though, replicate so quickly with such tight space constraints in their genomes (and are under such constant attack from their hosts) that a looser setting works out better for them in the long run. Even so, coronaviruses in general rather huge genomes already, perhaps the largest of the RNA viruses. A lot of that is devoted to proteins that affect the immune response of their hosts - which for SARS-CoV-2 now includes us, damn it all - and a lot of it codes for proteins involved in replication. Since viruses don't eat or mate, these are really where you'd expect the action to be anyway. The proteins that do the RNA synthesis and proofreading are nsp12 through nsp16, with nsp12 being the RNA polymerase itself and nsp14 being the proofreading enzyme. (See this overview of coronavirus biology for more). Those are the enzymes that have landed in the "good but not perfect" space, because a virus that went all the way up to nearly perfect fidelity could easily find itself boxed in as the environment around it changed, even if it were able to carry around all that high-quality enzymatic machinery at all. It works out better to throw the occasional mutation just to vary things up.

Bacteria are of course a lot larger and more complex than viruses, and as an aside, their "genetic fidelity" rheostat has a meta-rheostat of its own. Many bacteria switch over to more error-prone replication under stress, an adaptation that starts throwing Hail Mary passes in all directions. Obviously enough of these actually complete in the end zone for this strategy to have been handed down! Otherwise we'd never have heard of it again, naturally - we're seeing the descendents of the bacteria for which this last-ditch move paid off. Viruses don't have quite that level of control, though, so they end up at whatever single fidelity setting keeps the population going the best.

And as you'll see from that Stat article, that's more than enough to keep everyone working night and day. These mutations can potentially affect every job that a virus has on its list - how easily it enters human cells, how well it can evade host defenses along the way, how quickly it replicates, and more. It's important not to fall into teleological thinking when you look at these things: the only thing that counts is how well any given form of the virus does at making more virus. It doesn't have to necessarily change to become more deadly, for example, and in fact going too far along those lines actually will keepit from spreading as well as it could otherwise. Imagine a virus that drops people in their tracks, dead inside of a half hour, versus one that allow for days where a person can wander around spreading it. The second one has a real advantage in the Make More Virus competition, which is the only competition there is.

But evading a host's immune system could well have a side effect of making a viral infection worse, and the problem is that its more immediate effect is that it can Make More Virus. The same is obviously true for mutations that increase infectivity (which could happen via more effective binding to human cell membranes, or faster replication once that gets going, or any number of other things). We're already seeing the Delta variant showing up a clearly more infective, so the big questions are whether that could be ratcheted up even more, and whether that could also come along with increased evasion to the immune responses raised by either vaccination or prior infection with a slightly earlier coronavirus variant.

To a good approximation, the answer to those questions is that we have no idea (although it's not for lack of trying to figure them out). Over the last week or so, a lot of people have asked me about this preprint, which suggests that the current Delta variant may be a short distance away, mutationally, from one that could indeed do a better (well, worse for us) job of evading vaccine immunity. It's not at all a crank paper, and this is just the sort of question that has to be addressed - but the good news (from what I can see) is that this work is not as doomful as its title makes it seem. This Twitter thread goes into some of the reasons. A big one is that the cell line that's being used doesn't really recapitulate the cell-entry pathway used out here in the real world (it doesn't have the TMPRSS enzyme involved, to be specific). That's involved with the "furin cleavage site", which is a key part of SARS-CoV-2, and as Jeremy Kamil notes in that thread, viruses without the FCS can look very impressive when infecting things like Vero cells in vitro. But if you want to claim increased infectivity, you have to use a more realistic cell line, or (better yet) demonstrate in animal models.

He also says (correctly) that we truly have no idea what mutations will show up next, in what order, or on top of what existing variants. So it's not as meaningful as it sounds to say that "If you take Delta and add only X more changes you'll get a supervirus!" or whatever. The number of potential mutations in the coronavirus genome is beyond comprehension. There are 29,811 RNA residues in its sequence, and changing these produces a vast number of potential altered proteins. Amino acids can change, stop codons can show up, frameshifts can occur; it just goes on and on. Some of these are more likely than others, but even when we can go that far, their effects are almost entirely unpredictable. I would advise taking any "we're this close to a terrible variant" headines with a lot of sang-froid. I mean, we might be. Or we might not.

It's surely easier to get a handle on some protein surface region that we know to be important, rather than tackling the whole fitness landscape at once. This preprint does that by looking at computational modeling of all the amino acid residues at the Spike/RBD interface with human ACE2 (the protein binding event that that is the first step for viral entry). I'm not the greatest customer for this sort of approach - there are plenty of uncertainties in modeling protein-protein interactions - but the authors are especially looking for outliers, non-additive mutations that make a big change all by themselves. And they're not seeing much evidence for them, outside of the mutations we already have. The possibilities tend strongly to be incremental and additive, rather than Sudden New Landscape types, and they end by saying that ". . .the modest ensemble of mutations relative to the WT shown to reduce vaccine efficacy might constitute the majority of all possible escape mutations" So I will allow myself to be cheered up a bit by that, and I very much hope that they're right. I'm happy to imagine the coronavirus running out of options, but I'm not quite ready to break out the party hats just yet. The Pfizer scientists interviewed in Goodhill's article seem to be pretty evenly divided on the question, and honestly, there aren't any people in the world who are in a position to say more.

https://www.science.org/content/blog-post/keeping-coronavirus-variant-landscape